面向变工况迁移学习的泵阀小波残差网络异常检测

吴晓,张铭轩,何智烨,罗风,司国锦,夏唐斌

Anomaly detection using wavelet residual network for pump valves via transfer learning under variable operating conditions

Xiao WU,Mingxuan ZHANG,Zhiye HE,Feng LUO,Guojin SI,Tangbin XIA

表 4 不同模块对模型跨工况迁移平均准确率的影响分析

Tab.4 Analysis of impact of different modules on average accuracy of cross-condition transfer %

模型 Acc $\overline {\mathrm{Acc}} $ 工况1→1 工况1→2 工况1→3 工况1→4 TL-CWNet 92.79 91.10 92.49 92.59 92.24 ITL-CWNet 95.79 96.30 95.65 96.70 96.11 ITL-ICWNet 99.04 97.11 97.78 97.66 97.90 ITL-ICWMNet 99.90 99.15 99.54 99.85 99.61